Video-from-orbit payloads
Staring video sensors record 30–90 seconds of continuous footage from low Earth orbit, revealing motion that single frames cannot capture. The downlink cost is real; so is the intelligence advantage for traffic, plume and wake analysis.
Thirty seconds of staring changes the question you can answer
A panchromatic still image tells you a vehicle is present at a given coordinate at a given moment. A 30-second video clip at the same ground sampling distance tells you its speed, heading, whether it is accelerating, and whether other vehicles are reacting to it. That shift from presence to behaviour is the core proposition of video-from-orbit payloads, and it is not trivial.
The operational principle is straightforward. The satellite's attitude control system holds a fixed ground patch in the focal plane rather than scanning across it. SkySat, Planet's video-capable constellation, demonstrated this staring mode at roughly 1.1 m ground sampling distance and frame rates of around 30 frames per second, producing clips typically between 60 and 90 seconds before thermal and storage constraints intervene. The result is a short film of a few square kilometres of Earth, geocorrected and delivered as a video file. The physics are unforgiving: at 450–500 km altitude, an LEO platform moves at roughly 7.5 km/s relative to the ground, so the attitude system must compensate continuously to hold the stare. Any residual jitter blurs moving features and degrades the effective resolution below the optical limit.
The three use cases that actually justify the downlink budget
Video-from-orbit is expensive in downlink volume relative to a single frame. A 90-second clip at 30 fps is 2,700 individual images compressed into a sequence; even with aggressive compression, the data volume per tasking event is an order of magnitude larger than a comparable still. That cost is only worth bearing when the analytical question is inherently temporal.
Traffic flow analysis is the clearest case. Counting vehicles on a highway from a still is possible but ambiguous: parked vehicles and moving ones look identical. A 60-second clip allows speed estimation, lane-change detection and queue-length measurement at major junctions. Port and airport throughput studies use the same logic, tracking aircraft taxiing or vessels manoeuvring in confined water.
Plume analysis is the second validated use. Industrial stacks, wildfires and ship exhausts all produce plumes whose drift direction and dispersion rate encode wind speed and atmospheric stability at the emission altitude. A single frame gives you a snapshot of plume geometry. A video gives you a vector field. The distinction matters for emission attribution and for dispersion modelling used in emergency response.
Wake analysis is the third. A vessel moving through calm water leaves a Kelvin wake whose geometry is governed by hull speed, not heading. Video allows the wake to be tracked over tens of seconds, giving a speed estimate independent of any transponder signal. This is directly relevant to maritime domain awareness when a vessel has disabled its AIS.
Payload architecture: aperture, frame rate and the thermal ceiling
A video payload is essentially a high-frame-rate pushframe imager with a large focal plane array and a fast readout chain. The aperture determines ground sampling distance in the usual diffraction-limited way; a 35–50 cm aperture at 500 km yields roughly 1–2 m GSD in the visible band, which is the practical range for most current commercial video satellites. Larger apertures improve resolution but increase mass and thermal load, and the thermal load is the binding constraint for video mode specifically.
Continuous imaging at 30 fps generates heat in the detector and readout electronics at a rate that a small satellite's passive thermal system cannot sustain indefinitely. Published operational practice for SkySat-class payloads caps continuous video at around 90 seconds per pass. After that, the satellite typically reverts to still imaging while the focal plane cools. Mission planners should treat 60–90 seconds as the practical clip ceiling for a 100–200 kg class platform, not a specification to be negotiated away.
Storage and downlink are co-constraints. A single 90-second video clip at 1 m GSD can occupy 5–15 GB before compression depending on the bit depth and frame rate. Onboard compression (H.264 or similar) reduces this substantially, but the compressed file still competes with stills in the downlink queue. Programmes that task video heavily need ground station access sized for that throughput, not for a stills-only mission.
Where video fails, and why that matters before you commit
Cloud cover affects video exactly as it affects any optical payload. A 90-second stare over a cloud-covered target yields 90 seconds of cloud footage. Unlike SAR, there is no all-weather workaround. For time-sensitive monitoring of tropical ports or equatorial infrastructure, cloud probability at the target location should be modelled before video is specified as the primary collection mode.
Resolution floors are real. At 1–2 m GSD, individual pedestrians are not resolved; vehicle classification is possible but not reliable for all vehicle types. Sub-metre video from orbit exists in limited form but requires larger apertures and correspondingly larger platforms, which raises cost and revisit trade-offs substantially.
Revisit is the sharpest limit. A single video-capable satellite over a given target delivers one 90-second clip per pass, with pass opportunities typically once or twice per day at best for a sun-synchronous orbit. If the event of interest (a convoy departure, a vessel getting under way) does not coincide with the pass window, it is missed. Constellation size directly controls this risk, and building or procuring a dedicated video constellation is a different programme scale from a single satellite.
Finally, video data is harder to exploit automatically than stills. Change detection pipelines built for still imagery do not transfer directly to video. Motion extraction, object tracking and speed estimation require separate processing chains. Governments acquiring a video-capable satellite should budget for the analytics infrastructure, not just the downlink.
Fitting video into a sovereign programme architecture
Video capability is almost never the primary payload on a sovereign first satellite. The downlink burden, the thermal constraints and the specialised analytics pipeline make it a second-mission or supplementary-payload decision in most national programmes. The more common architecture is a high-resolution optical primary with a video mode added to the focal plane design, accepting the trade that video clips will be shorter and less frequent than a dedicated video satellite would deliver.
Where video does appear as a primary requirement, it is usually driven by a specific operational need: port monitoring, border crossing surveillance or disaster response where the temporal dimension of the event is the intelligence. In those cases, the programme should be sized around the downlink and ground processing requirements from the outset, because retrofitting throughput capacity after launch is expensive and sometimes impossible.
Satellize's crop-estimation programme for the Kingdom of Tonga uses multispectral stills rather than video, which reflects the agronomic question being asked. Vegetation indices do not require motion. That distinction is the right way to think about video: it is the answer to a specific class of question, not a general upgrade to an optical programme.
Engineering parameters
| Typical platform mass class | 80–250 kg (SkySat-class); sub-50 kg video payloads exist but at coarser GSD |
| Ground sampling distance (visible) | 0.9–3 m depending on aperture and altitude; sub-metre video remains rare and expensive |
| Frame rate | 25–30 fps typical; some systems up to 60 fps at reduced resolution |
| Maximum continuous clip duration | 60–90 seconds per pass for 100–200 kg platforms; thermally constrained |
| Raw data volume per clip (pre-compression) | 5–20 GB for a 90-second clip at 1 m GSD; compressed delivery typically 500 MB–3 GB |
| Payload power draw (imaging mode) | 40–120 W depending on focal plane size and readout speed |
| Spectral band | Panchromatic (450–900 nm) standard; some systems add narrow RGB channels |
| Revisit (single satellite, SSO) | 1–2 passes per day over a given latitude; event capture depends on pass timing |
| Downlink requirement | X-band at 300–800 Mbps typical; high-volume tasking requires dedicated ground station capacity |
| Geolocation accuracy | 5–50 m CE90 without ground control points; 2–5 m with GCPs depending on attitude knowledge |
One contract, one accountable engineer
Commissioned as one programme, not a stack of contracts: spacecraft, launch, ground segment, mission control, training and handover are priced together. Source-access terms and audit rights are agreed in writing before signature. Discuss video payload trade-offs for your mission.